Scientific research special disease library system based on hepatolenticular degeneration

Abstract

The invention discloses a special scientific research disease library system based on hepatolenticular degeneration, which belongs to the field of medical information technology and clinical scientific research and comprises a case data acquisition module, a database management module, a clinical scientific research analysis module, a follow-up management module and a standardized quality control module. Structured input of copper metabolism indexes, clinical manifestation, genetic information and treatment data is realized through a disease specificity data acquisition template; semantic fusion and efficient retrieval of multi-source heterogeneous data are completed by utilizing a database architecture based on a knowledge graph; a copper metabolism kinetic model, a nervous system curative effect evaluation model and a genotype-phenotype association model are embedded, so that automatic analysis of scientific research data is realized; and generating a personalized follow-up plan and dynamically adjusting the follow-up period in combination with a rule engine. The system realizes data acquisition standardization, scientific research analysis intelligence and follow-up management systematization, and is suitable for high-quality accumulation and comprehensive research of hepatolenticular degeneration scientific research data.

Description

Technical Field

[0001] This invention belongs to the field of medical information technology and clinical research, and in particular relates to a research disease database system based on Wilson's disease. Background Technology

[0002] Wilson's disease (Wilson's disease) is an autosomal recessive inherited disorder of copper metabolism caused by a mutation in the ATP7B gene. Its pathological feature is the abnormal accumulation of copper ions in multiple organs, including the liver, brain, cornea, and kidneys, leading to liver dysfunction, neurological disorders, and multi-system complications. In recent years, with the development of precision medicine and big data analytics, clinical researchers have begun to utilize general medical research systems or medical big data platforms to conduct etiological studies and clinical analyses of Wilson's disease. These platforms typically integrate electronic medical records, laboratory tests, and imaging data to achieve centralized management of research data. However, these systems are generally based on a general design and lack specific support for the pathological characteristics of Wilson's disease, dynamic indicators of copper metabolism, genetic variation patterns, and long-term follow-up data, making it difficult to directly meet the needs of refined and longitudinal disease research.

[0003] Existing technologies still have the following problems in the research context of Wilson's disease:

[0004] First, data collection is fragmented and lacks standardized criteria. Different hospital systems lack unified data element definitions for core copper metabolism indicators and clinical manifestations, making it difficult to integrate and compare data.

[0005] Secondly, data management still relies mainly on manual entry or general databases, lacking automated quality control mechanisms, which easily leads to entry errors and data loss.

[0006] Third, existing analytical tools fail to embed specialized models corresponding to disease mechanisms, requiring researchers to invest significant manpower in data cleaning, format conversion, and model building, thus limiting both the depth and efficiency of the analysis.

[0007] Fourth, the follow-up management system is imperfect, lacking rule-based personalized plans and dynamic adjustment mechanisms, resulting in long-term data gaps and loss of research samples.

[0008] Therefore, there is an urgent need for a research-specific disease database system targeting the characteristics of Wilson's disease, which should be systematically optimized in terms of standardized data collection, knowledge graph-based management, disease-specific analysis modeling, and intelligent follow-up, so as to achieve high-quality accumulation and in-depth application of research data. Summary of the Invention

[0009] To address the aforementioned technical problems, this invention provides a research-specific disease database system based on Wilson's disease, comprising:

[0010] The system includes a case data collection module, a database management module, a clinical research analysis module, a follow-up management module, and a standardized quality control module.

[0011] The case data acquisition module obtains structured core indicators of copper metabolism, clinical manifestations, genetic information and treatment data based on disease-specific templates;

[0012] The database management module constructs a knowledge graph containing semantic relationships between gene mutations, clinical phenotypes, treatment plans, and detection indicators, enabling unified storage and efficient retrieval of multi-source data.

[0013] The clinical research analysis module analyzes copper metabolism dynamics, non-invasive assessment of liver fibrosis, neurological efficacy, and genotype-phenotype relationships by embedding specialized analysis models.

[0014] The follow-up management module automatically generates and dynamically adjusts personalized follow-up plans based on patient classification, severity, and treatment plan.

[0015] The standardized quality control module performs logical, reasonable, and complete checks based on knowledge graph rules, and continuously manages the quality of the data.

[0016] Optionally, the case data acquisition module interfaces with the hospital information system, laboratory information system and image archiving system to obtain data such as serum ceruloplasmin, 24-hour urinary copper, liver copper content and UWDRS score according to preset field mapping rules, and performs speech transcription and standardization processing on unstructured text information.

[0017] Optionally, the database management module includes:

[0018] The entity recognition unit is used to extract entities such as gene mutations, clinical phenotypes, and detection indicators from clinical texts, literature, and medical records based on the BERT-BiLSTM-CRF model.

[0019] A relation extraction unit is used to determine the semantic relationships between the entities using a Span-Pair classification model.

[0020] The fusion processing unit is used to align synonymous entities from different sources based on the edit distance algorithm and the Wu-Palmer semantic similarity fusion rule, and generate a unified node identifier.

[0021] Optionally, the database management module uses the Neo4j graph database to store the knowledge graph structure, defining each entity as a node and semantic relationships as edges.

[0022] The relationship weight is determined by calculating the co-occurrence frequency between nodes, and the priority path is sorted in multi-hop path query according to the weight.

[0023] The system evaluates the credibility of the query results based on weight thresholds and indicates the strength of the relationship using a color gradient in the results visualization interface.

[0024] Optionally, the clinical research analysis module includes a copper metabolism kinetics analysis submodule.

[0025] This submodule fits time-series data of serum ceruloplasmin and urinary copper concentrations to a nonlinear mixed-effects model.

[0026] Based on the individualized parameter estimation model, the synthesis rate constant and the clearance rate constant are used to calculate the time point at which steady state is reached using the maximum likelihood algorithm.

[0027] The model results were then correlated with liver function indicators and neurological symptom scores.

[0028] Optionally, the clinical research analysis module further includes a sub-module for evaluating the efficacy of treatment in the nervous system.

[0029] This submodule analyzes the impact of different copper-chelating drug treatment regimens on the trend of UWDRS score changes using a linear mixed-effects model.

[0030] Interactive items are established based on patient treatment group, time point, and baseline score, and a visual efficacy trend chart and statistical report are automatically generated after the analysis is completed.

[0031] Optionally, the follow-up management module includes a rule base, an inference engine, and an execution unit.

[0032] The rule base defines follow-up conditions and inspection items in Boolean logic form, and the inference engine matches patient classification, UWDRS score and treatment plan based on the Rete algorithm.

[0033] When the monitoring data meets the preset trigger conditions, the execution unit calls the update function to automatically adjust the follow-up interval and detection items, and writes the update record into the database.

[0034] Optionally, the standardized quality control module includes a rule definition unit, a verification execution unit, and a quality reporting unit.

[0035] The rule definition unit sets three types of rules: logical consistency, reasonableness, and completeness, which are used to judge the logical consistency between fields, the reasonableness of the numerical range, and the completeness of key data, respectively.

[0036] The verification execution unit performs rule verification during the data entry stage and the periodic batch detection stage, and generates an anomaly report with location information.

[0037] The quality reporting unit statistically analyzes the completeness, consistency, and reasonableness of data, and updates the rule base based on the PDCA cycle mechanism to achieve continuous improvement.

[0038] Compared with the prior art, the present invention has the following advantages and technical effects:

[0039] This invention constructs a Wilson's disease-specific database system, employing disease-specific data collection templates, a knowledge graph-based database architecture, embedded specialized research analysis models, and a rule-based personalized follow-up mechanism. This achieves structured, intelligent, and traceable management of the entire research process. The system eliminates data fragmentation and non-standardization at the source, significantly improving data entry accuracy and query efficiency. By embedding copper metabolism kinetics and genotype-phenotype association models, it enables in-depth mining and automated analysis of high-dimensional research data. Through a rule-driven dynamic follow-up plan, it ensures the continuity and integrity of patient longitudinal data, effectively solving problems such as data silos, superficial analysis, and disconnected follow-up in existing technologies. This creates a high-quality, integrated information platform for Wilson's disease research applications. Attached Figure Description

[0040] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0041] Figure 1 This is a schematic diagram of a module according to an embodiment of the present invention. Detailed Implementation

[0042] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0044] Example 1

[0045] like Figure 1 As shown, this embodiment provides a research-specific disease database system based on Wilson's disease, including:

[0046] The system includes a case data collection module, a database management module, a clinical research analysis module, a follow-up management module, and a standardized quality control module.

[0047] The case data acquisition module obtains structured core indicators of copper metabolism, clinical manifestations, genetic information and treatment data based on disease-specific templates;

[0048] The database management module constructs a knowledge graph containing semantic relationships between gene mutations, clinical phenotypes, treatment plans, and detection indicators, enabling unified storage and efficient retrieval of multi-source data.

[0049] The clinical research analysis module analyzes copper metabolism dynamics, non-invasive assessment of liver fibrosis, neurological efficacy, and genotype-phenotype relationships by embedding specialized analysis models.

[0050] The follow-up management module automatically generates and dynamically adjusts personalized follow-up plans based on patient classification, severity, and treatment plan.

[0051] The standardized quality control module performs logical, reasonable, and complete checks based on knowledge graph rules, and continuously manages the quality of the data.

[0052] The system's hardware environment includes: a central server employing a high-performance rack-mount server (CPU: 2x Intel Xeon Gold 6348, Memory: 256GB DDR4 ECC, Storage: 2x 3.84TB NVMe SSD + 4x 16TB 7.2KRPM SAS HDD), deployed in a data center meeting Level 3 cybersecurity standards. The network adopts a 10 Gigabit core switch and Gigabit-to-the-desktop architecture, and deploys a next-generation firewall, WAF, and intrusion detection system.

[0053] Software environment: The server operating system is Ubuntu Server 20.04 LTS; the database uses a MySQL 8.0 cluster (one master and two slaves); the backend business microservices are developed using the Java Spring Boot 2.7 framework; the frontend user interface is built using Vue.js 3 + Element Plus; the data analysis service is based on Python 3.9 and integrates libraries such as Pandas, NumPy, SciPy, Statsmodels, Scikit-learn, Matplotlib, and Seaborn.

[0054] During system initialization, the administrator imports a predefined data acquisition template specific to Wilson's disease (containing 12 major categories and over 200 structured fields), standardized process specification documents, and a medical terminology dictionary based on SNOMED CT and LOINC via a web interface. Database backup strategies are configured (incremental backups every 15 minutes, a full backup daily at 02:00, and a synchronization delay of less than 1 minute at the off-site disaster recovery center) and user role permissions (e.g., principal investigator, research physician, data entry clerk, auditor, each with different data access and operation permissions).

[0055] Specifically, this includes the case data collection module:

[0056] Key innovations: Disease-specific data templates and multi-source intelligent access. This module pre-defines a structured data acquisition template specific to Wilson's disease. This template is not a simple field extension of a general template, but rather, based on domestic and international treatment guidelines and expert consensus, it mandates a complete data element system strongly relevant to Wilson's disease diagnosis, treatment, and research. Its core includes:

[0057] Key indicators of copper metabolism: serum ceruloplasmin (units are mandatory to be uniformly expressed as mg / dL), 24-hour urinary copper (units are mandatory to be uniformly expressed as μg / 24h), serum copper (μg / dL), and liver copper content (μg / g dry weight, applicable to patients with liver biopsy data).

[0058] Structured fields for clinical manifestations: neurological symptoms and signs (structured entry using UWDRS scores), corneal KF rings (standardized selection descriptions divided into grades 0-IV, which can be linked to images), and liver manifestations (Child-Pugh classification, liver elasticity stiffness value in kPa).

[0059] Genetic information module: ATP7B gene mutation site, mutation type (missense, nonsense, splicing, etc.), homozygous / complex heterozygous status.

[0060] Treatment regimen and adverse reactions: Detailed records of the start time, dosage, and course of treatment for specific drugs such as penicillamine, sodium dimercaptopropanesulfonate, and zinc preparations; and structured records of adverse drug reactions, such as proteinuria and granulocytopenia caused by penicillamine.

[0061] This module supports intelligent access to multi-source data: it interfaces with hospital HIS, LIS, and PACS systems through configured standardized interfaces (such as HL7, FHIR, and DICOM) to automatically capture and map the aforementioned structured data. For unstructured information that cannot be automatically obtained (such as detailed medical history texts), a convenient manual input interface is provided, and voice input to text conversion is supported. The system has a built-in terminology standardization service, based on standard medical dictionaries such as SNOMED CT, LOINC, and ICD-10, to perform real-time normalization processing on the entered content to ensure data consistency.

[0062] Database management module:

[0063] Key innovation: In one embodiment, the database management module achieves storage and efficient retrieval by constructing a knowledge graph specific to Wilson's disease. The construction process includes:

[0064] Data processing and entity recognition;

[0065] The first step in knowledge graph construction is data collection and processing. This system focuses on Wilson's disease and has collected heterogeneous data from multiple sources, including medical literature, clinical records, and gene databases. This data contains rich information, but its diverse formats and complex structures require preprocessing to transform it into a form suitable for subsequent analysis.

[0066] In the entity recognition stage, the system uses the BERT-BiLSTM-CRF model. The model structure is as follows:

[0067] 1. BERT Layer: As a pre-trained language model, BERT encodes input text using a bidirectional Transformer architecture, effectively capturing the complex relationships between each word and its surrounding context, generating high-quality contextual representations for each word. For example, in the text describing the symptoms of Wilson's disease, "the patient experienced limb tremors and slurred speech," BERT can fully understand the semantic features of "limb tremors" and "slurred speech" in that context.

[0068] 2. BiLSTM Layer: The feature vector output by BERT is input into a bidirectional long short-term memory (BiLSTM) network. BiLSTM can process sequence data in two directions (from front to back and from back to front), making full use of contextual information to more accurately model the position and role of each word in the entire sequence, and further extract rich feature representations. For example, when analyzing the sentence "ATP7B gene mutation leads to abnormal copper metabolism", BiLSTM can accurately grasp the features of entities such as "ATP7B gene", "mutation", and "abnormal copper metabolism" by combining the context.

[0069] 3. CRF Layer: Conditional Random Field (CRF) is used for global sequence labeling. Its scoring function is S(X,Y)=\sum_{i =1}^{n}(W_{y_i} \cdot x_i+T_{y_{i - 1},y_i}), where X is the input sequence, Y is the label sequence, n is the sequence length, W_{y_i} is the weight vector associated with label y_i, x_i is the input feature vector, and T_{y_{i - 1},y_i} is the transition score from label y_{i - 1} to y_i. CRF considers the global dependencies of the labeled sequence. When predicting the label of each character, it not only depends on the features of the current character but also comprehensively considers the labeling of its neighboring characters. Finally, the Viterbi algorithm is used to decode the optimal label sequence from the output of the CRF layer to obtain the entity recognition result. The Viterbi algorithm is based on the principle of dynamic programming. It uses the recursive formula dp(t, s) = P(O_t|s)\max_{s'} (dp(t - 1, s')P(S_{t}|s')) (where dp(t, s) represents the probability of the most likely path to state s with observation O_t at time t, P(O_t|s) is the observation probability, and P(S_{t}|s') is the state transition probability) to find the sequence with the highest probability among all possible label sequences, thereby identifying entities in the text, such as gene mutations, clinical phenotypes, and laboratory indicators.

[0070] Relation extraction;

[0071] Entity relation extraction aims to identify semantic relationships between identified entities. This system employs a Span-Pair classification model for relation extraction. This model first identifies entities from the NER layer using specific boundaries and type-specific characters, then processes each possible entity pair. Specifically, for each entity pair, the encoding of its first token is concatenated, then connected to a fully connected layer, and finally, a softmax algorithm is used to classify and determine the relationships between the entity pairs, identifying relationships such as "cause," "manifest as," and "decline after treatment." For example, for the text "ATP7B gene mutation causes disease," the model can accurately identify the "cause" relationship between "ATP7B gene mutation" and "cause disease."

[0072] Knowledge integration;

[0073] During the knowledge graph construction process, inconsistencies in entity naming may arise from different data sources. To address this issue, the system employs a strategy that combines edit distance and Wu-Palmer semantic similarity for entity alignment and conflict resolution.

[0074] 1. Edit Distance Calculation: String similarity is calculated using Sim_edit(s_1, s_2) = 1-\frac{Lev(s_1, s_2)}{\max(|s_1|, |s_2|)}, where Lev(s_1, s_2) is the Levenshtein distance between s_1 and s_2, representing the minimum number of single-character editing operations (insertion, deletion, replacement) required to transform one string into another. For example, for the similar entity names "Copper Blue Protein" and "Copper Blue Protein", the degree of similarity can be measured by calculating the edit distance.

[0075] 2. Semantic Similarity Calculation: Semantic similarity is calculated using the Wu-Palmer formula supported by the ontology structure: Sim_WUP(c_1, c_2) = \frac{2 \cdot depth(LCS(c_1, c_2))}{depth(c_1)+depth(c_2)}, where c_1 and c_2 are concepts in the ontology, depth(LCS(c_1, c_2)) is the depth of the Least Common Subsumer (LCS) of c_1 and c_2 in the ontology structure, and depth(c_1) and depth(c_2) are the depths of c_1 and c_2 in the ontology structure, respectively. This formula measures the similarity between two concepts from a semantic level based on ontology structure information.

[0076] 3. Comprehensive Judgment: Edit distance similarity and semantic similarity are fused to obtain a comprehensive score. When the comprehensive score is higher than the threshold (0.85), the two entities are considered to be the same entity. For conflicting attributes, majority voting and data source priority are used for processing. For example, if most data sources consider the mutation frequency of a certain gene to be a specific value, while a few data sources have different values, the value of the majority data source is adopted; if different data sources have the same number of votes, the selection is based on the pre-set data source priority.

[0077] Knowledge modeling;

[0078] The system constructed a dedicated medical ontology for Wilson's disease to define the knowledge structure of this domain. Details are as follows:

[0079] 1. Defining Entities and Relationships: Seven types of entities were identified, including gene mutations, treatment regimens, and UWDRS scores, as well as their semantic relationships, such as "leads to" and "decreases after treatment." These entity and relationship definitions are based on a deep understanding and analysis of knowledge in the field of Wilson's disease, ensuring accurate representation of key information within this domain.

[0080] 2. Ontology Construction Tool: The Protégé tool was used to construct the OWL format ontology. Protégé is a widely used ontology editing tool that provides a visual interface, facilitating the creation, editing, and management of ontology by domain experts and knowledge engineers. With Protégé, ontology elements such as classes, attributes, and individuals can be defined intuitively, and relationships and constraints between them can be set. The constructed OWL format ontology possesses strong semantic expressiveness and logical reasoning capabilities, supporting subsequent logical reasoning and graph queries. For example, reasoning can be performed based on relationships defined in the ontology, such as inferring the indirect association between a treatment regimen and a gene mutation from the relationship between "a gene mutation leads to a certain clinical phenotype" and "a treatment regimen targets this clinical phenotype."

[0081] Graph storage;

[0082] The knowledge graph is stored in the Neo4j graph database, and its specific storage method and characteristics are as follows:

[0083] 1. Storage Structure: In Neo4j, nodes represent entities, and edges represent relationships between entities. For example, the node "ATP7B gene mutation" is connected to the node "cause" via the edge "leads to," clearly demonstrating the semantic association between entities. This graph structure can intuitively represent complex relationships in a knowledge graph, facilitating querying and analysis.

[0084] 2. Relationship Weight Calculation: Relationship weights are calculated based on co-occurrence frequency, using the formula w = log(f / N - f + 1), where f is the number of times the entity pair (mutation and phenotype) co-occurs, and N is the total number of samples. For example, if 20 out of 100 samples simultaneously show "ATP7B gene mutation" and "pathogenicity," then f = 20 and N = 100. The weight of the relationship between them can be calculated using this formula. Relationship weights are used for path optimization and credibility ranking. When performing multi-hop path queries, relational paths with higher weights are more likely to be prioritized. Weights also serve as an important reference when evaluating the credibility of query results; higher weights indicate higher credibility of the relationship. Neo4j supports multi-hop path queries and semantic reasoning, efficiently handling complex query requests, such as querying "compound heterozygous mutation, clearly diagnosing Wilson's disease," and performing semantic reasoning based on the stored knowledge graph structure and relationships to uncover potential knowledge associations.

[0085] Atlas applications;

[0086] The knowledge graph supports genotype-phenotype association analysis, personalized follow-up plan generation, and data logic verification.

[0087] Preferably, the relation weights are calculated using statistical correlation. Let f be the co-occurrence frequency of a mutation and a phenotype in the sample, and N be the total number of samples. Then the weight w satisfies: w = log[(f / (N−f))+1].

[0088] This weight is used for path selection and credibility ranking during subsequent reasoning.

[0089] Although entity recognition and relation extraction can be accomplished using existing deep learning models, the ontology structure, entity types, relation definitions, and weight functions are all specifically designed for the field of Wilson's disease, and are not a simple transfer from general medical knowledge graphs.

[0090] A distributed hybrid storage architecture is adopted, with hot data (recent active cases) stored on SSDs and cold data (historical archived data) stored on HDDs. The database schema design deeply integrates clinical knowledge of Wilson's disease, establishing specific semantic relationships between entities to form a small disease-specific knowledge graph. For example, the entity of "gene mutation site" is strongly associated with the entities of "clinical phenotype" (hepatic / neurological / mixed) and "treatment response," and an "indicator-time" sequence relationship is established, providing a direct and efficient data query foundation for studying the genotype-phenotype-prognostic association.

[0091] It provides a multi-condition combined search function, supporting precise or fuzzy queries on all the above-mentioned specific fields, such as "finding all patients with ATP7B gene p.R778L site mutation, accompanied by severe neurological symptoms (UWDRS>50), and treated with sodium dimercaptopropanesulfonate for more than 1 year", with the search response time optimized to within 3 seconds.

[0092] Establish a three-tiered data security backup and disaster recovery mechanism: real-time incremental backup (every 15 minutes), daily full backup, and cross-geographical disaster recovery (data is encrypted and synchronized via a dedicated network). Implement a role-based access control (RBAC) system to finely control user permissions for adding, deleting, modifying, querying, and exporting data. All operations are logged, and log audits are maintained for at least 5 years.

[0093] Clinical research analysis module:

[0094] Core innovation: Embedded algorithm set specifically for Wilson's disease analysis. This is the core value of this invention. This module integrates several pre-set analytical models for the core scientific questions of Wilson's disease that are not available in general statistical software.

[0095] Copper metabolism kinetics analysis model: Based on time series data, this model analyzes the dynamic changes of serum ceruloplasmin, urinary copper, and other indicators in individual patients or groups undergoing copper chelation therapy, predicts the time to reach a steady state, and performs correlation analysis with prognosis (such as improvement in liver function and relief of neurological symptoms). In one embodiment, the copper metabolism kinetics analysis model is used to predict the dynamic changes of serum ceruloplasmin and urinary copper during copper chelation therapy. The model adopts a nonlinear mixed-effects structure. Let Cp(t) be the serum ceruloplasmin concentration at time t, Cp0 be the baseline value, k_in be the synthesis rate constant, k_out be the clearance rate constant, η_i be the individual random effect, and ε_ij be the residual error. Then:

[0096] lnCp(t)=lnCp0·e^(−k_out·t)+(k_in / k_out)·(1−e^(−k_out·t))+η_i+ε_ij;

[0097] The model parameters were obtained by fitting the model using the maximum likelihood estimation method and were used to predict the time to reach steady state and its association with the improvement of liver function and neurological symptoms.

[0098] Association analysis for non-invasive assessment of liver fibrosis: This study specifically analyzes the quantitative relationship between biochemical indicators such as serum ceruloplasmin and urinary copper and the value of liver transient elastic stiffness test (FibroScan), establishes a regression model, and attempts to use readily available biochemical indicators to assist in the assessment of the degree of liver fibrosis.

[0099] Neurological Efficacy Evaluation Model: Based on UWDRS score data from multiple follow-up visits of patients, repeated measures ANOVA or a mixed-effects model is used to quantitatively evaluate the rate of improvement and final effect of different treatment regimens (or different doses of the same regimen) on neurological symptoms, and can generate individualized efficacy trend charts. In one embodiment, the neurological efficacy evaluation model is used to quantify the impact of different copper chelation regimens on UWDRS scores. A linear mixed-effects model is used, where Yi_ij is the UWDRS score of the i-th patient at the j-th follow-up visit, T_ij is time, Treatment_i is the treatment regimen, u_i is the individual random intercept, and ε_ij is the residual, then:

[0100] Y_ij=β0+β1·T_ij+β2·Treat_i+β3·(T_ij·Treat_i)+u_i+ε_ij;

[0101] The model estimates the trend of treatment effect over time and the differences between groups, and generates individualized efficacy curves for clinical decision-making reference.

[0102] Genotype-phenotype association analysis tool: Automatically integrates genetic and clinical data, performs chi-square tests, Fisher's exact tests, or logistic regression to analyze the statistical association between specific gene mutation types and clinical manifestations (age of onset, affected organs, disease severity). In one embodiment, the genotype-phenotype association analysis model is used to assess the statistical association between a specific ATP7B mutation and clinical manifestations. Let the genotype matrix G∈{0,1,2}^n×p, and the phenotype vector Y∈{0,1}^n, using L1 regularized logistic regression:

[0103] min_β{Σ_{i=1}^nlog[1+e^(−y_i(β0+β^Tg_i))]+λ||β||1};

[0104] Where λ is the regularization parameter and β is the regression coefficient vector, the model outputs the effect strength and significance level of each mutation site on the phenotype, which is used to screen high-risk mutations and guide accurate typing.

[0105] Meanwhile, the module provides a visual analysis interface that automatically converts analysis results into publication-ready charts (such as line charts, heatmaps, and forest plots), and supports one-click generation of analysis reports in Word or PDF format containing statistical details.

[0106] Follow-up management module:

[0107] Key innovation: Automatic generation and dynamic adjustment of rule-based personalized follow-up plans. This module incorporates a Wilson's disease follow-up rule engine, with its rule base defined by clinical experts. The system can automatically recommend and generate structured personalized follow-up plans based on the patient's disease type (hepatic / neurological / mixed), severity (e.g., Child-Pugh classification, UWDRS baseline score), and initial treatment plan at initial diagnosis. In one embodiment, the follow-up rule engine includes a rule base, working memory, a forward chain inference mechanism, and an execution module. The rule base is predefined by clinical experts in Boolean logic form, for example: if patient type = "neurological" ∧ UWDRS>50 ⇒ follow-up interval = "1 month" ∧ examination items = {UWDRS, urinary copper, ALT}. The inference mechanism uses a Rete-type forward chain algorithm to match patient factual data with rule conditions in real time; the execution module converts successfully matched rules into structured follow-up plans and writes them to the database. When abnormal indicators (such as ALT>120 U / L) occur during the follow-up, the system will trigger inference again, automatically upgrade the follow-up frequency to "weekly" and add tests such as coagulation function and bilirubin, so as to achieve dynamic adjustment.

[0108] Example rule: For patients with "acute liver failure", an intensive follow-up plan of "weekly monitoring of liver function and coagulation function" is automatically generated; for patients with "stable neurological type", a routine plan of "monitoring urinary copper every 3 months and performing UWDRS score and liver ultrasound every 6 months" is generated.

[0109] The system offers multi-channel automatic reminders (SMS, email, and in-system messages) to remind patients and researchers of follow-up appointments in advance. During follow-ups, doctors can conveniently enter data on mobile devices or computers. The system automatically links the data from this follow-up visit with the patient's baseline and historical follow-up data to create a complete longitudinal medical record. It also features automatic alerts for abnormal values ​​(such as a sudden drop in urinary copper, which may indicate sampling errors or compliance issues; or a sharp increase in liver enzymes, which may indicate potential liver damage), triggering alerts for doctor intervention.

[0110] Standardization and Quality Control Module:

[0111] Core innovation: Logical verification and continuous quality improvement based on medical knowledge graph. In one embodiment, the logical verification module establishes three types of verification rules based on the disease-specific knowledge graph: (1) Logicality rule: Let KF ring positive be a Boolean variable K, and serum ceruloplasmin be C (mg / dL), then if K=True ∧ C>20 ⇒ trigger logical abnormality; (2) Reasonableness rule: Let urinary copper be U (μg / 24h), and treatment status be T, then if U>1000 ∧ T=False ⇒ trigger reasonableness warning; (3) Integrity rule: Let the key field set R, if ∃r∈R is null ⇒ trigger integrity error.

[0112] The system generates monthly data quality reports, calculating completeness, consistency, and reasonableness rates, and continuously improves quality based on the PDCA closed-loop rule base. This module acts as a "gatekeeper" for ensuring the quality of disease-specific data. It possesses a data quality rule base specific to Wilson's disease, defined based on the clinical pathophysiology of this disease, and can perform real-time batch verification during and after data entry.

[0113] Logical validation: For example: "If the 'corneal KF ring' field is 'positive', then the 'serum ceruloplasmin' field should not be blank or higher than the normal value"; "The patient is less than 3 years old, but 'cirrhosis' is recorded, which indicates a logical error."

[0114] Reasonableness verification: For example: "The serum ceruloplasmin test result is >50mg / dL, which is outside the range of common Wilson's disease patients. Please confirm whether the unit is mg / dL"; "24-hour urinary copper excretion is >1000μg / 24h, which is consistent with the expected response to copper chelation therapy. If no treatment is given, it suggests a possible diagnosis."

[0115] Integrity check: Enforce that key fields (such as age at initial diagnosis, symptoms at initial diagnosis, and serum ceruloplasmin) must not be empty.

[0116] The system automatically generates data quality assessment reports periodically (e.g., monthly), calculates indicators such as data integrity rate, accuracy rate, and standardization rate, and pinpoints the location of specific problematic data to guide data administrators in cleaning and correction, thus forming a closed-loop quality management system.

[0117] Beneficial effects:

[0118] Compared with the prior art, the system provided in this embodiment produces the following significant beneficial effects, which directly stem from the aforementioned disease-specific innovative design:

[0119] It has achieved in-depth standardization and structuring of scientific research data: through disease-specific data templates and terminology standardization services, it ensures from the source that the collected data is highly structured, machine-readable, and directly serves the research on Wilson's disease, completely solving the problems of data fragmentation and non-standardization, and laying a solid foundation for high-quality research.

[0120] It significantly improves data management efficiency and security: the knowledge graph-based storage architecture and efficient retrieval functions enable researchers to accurately locate research queues within seconds, reducing the time required from hours or even days in the traditional method to seconds. A robust security backup and access control system ensures the security and compliance of sensitive medical data, greatly reducing management risks and costs.

[0121] It provides powerful and dedicated scientific research analysis capabilities: the embedded disease-specific analysis models (such as copper metabolism kinetic analysis and genotype-phenotype association analysis) can directly and efficiently answer the core scientific questions in Wilson's disease research, saving researchers the tedious and error-prone data preprocessing, transformation and basic modeling process from raw data to analysis results, improving analysis efficiency by several orders of magnitude and lowering the biostatistics threshold.

[0122] This ensures the continuity and scientific rigor of follow-up data: Personalized follow-up plans based on clinical rules guarantee the systematic and comparable nature of data collection. Automatic reminders and convenient data entry functions significantly improve patient compliance and data entry efficiency. Seamless correlation between follow-up data and baseline data constitutes a complete longitudinal dataset, greatly enhancing the value of disease progression research and prognostic factor analysis.

[0123] This constitutes a holistic solution to the research challenges of Wilson's disease: the collaborative workflow between modules (from specific data collection and intelligent management to in-depth analysis and dynamic follow-up) forms an organic and highly integrated whole, producing a technical effect of "1+1>2". This is not a simple patchwork or application of existing Class A or Class B technologies, but an innovative, non-obvious, systematic solution specifically designed to overcome the unique challenges in Wilson's disease research.

[0124] Example 2

[0125] This embodiment provides a research-specific disease database system based on Wilson's disease, including:

[0126] Take, for example, a newly diagnosed patient with Wilson's disease, "Zhang." Researchers create new cases in the system.

[0127] Automatic data acquisition: The system automatically retrieves Zhang's serum ceruloplasmin (12.5 mg / dL) and urinary copper (350 μg / 24h) test results via the LIS interface. It also retrieves thumbnails and report links for his liver ultrasound and cranial MRI images via the PACS interface.

[0128] Manual entry: In the structured interface, the researcher selects to enter the clinical manifestations: neurological symptoms - "tremor" and "dystonia", and selects the UWDRS score of 45 from the drop-down menu; corneal KF ring - select "Grade III"; gene test results - enter "p.R778L / p.P992L" in the ATP7B gene field.

[0129] Real-time quality control: During data entry, the system verifies the logic in real time. Since the KF ring was positive and serum ceruloplasmin was significantly reduced, the system passed the verification. However, when attempting to mistakenly enter urinary copper units as "mg / 24h," the system immediately displayed a warning: "Unit suspected to be incorrect; the common unit is μg / 24h. Please confirm!" This prevented serious data errors.

[0130] Example 3

[0131] This embodiment provides a research-specific disease database system based on Wilson's disease, including:

[0132] Specific scientific research analysis examples;

[0133] A research team wanted to study the "effects of different copper chelation regimens on improving neurological symptoms." After logging into the system:

[0134] Cohort construction: In the advanced search interface, set the search criteria: Diagnosis = "Wilson's disease" AND Main clinical manifestations = "Neurological type" AND Baseline UWDRS score > 30 AND Follow-up times >= 3. The system returns a cohort containing 85 patients within 2 seconds.

[0135] Analysis Execution: Researchers selected "Neurological Efficacy Evaluation Model" in the "Clinical Research Analysis Module". They selected "Treatment Regimen" as the grouping variable (Group A: Penicillamine, Group B: Sodium Dimercaptopropanesulfonate) and "UWDRS Score" as the outcome variable, with time points at baseline, 3 months, 6 months, and 12 months. Then, they clicked "Analyze".

[0136] Results Acquisition: The system automatically calls the Linear Mixed Model for analysis, handles missing values, and generates results.

[0137] Visualization results: A trend line graph (with confidence intervals) showing the change of UWDRS scores over time in the two groups of patients clearly shows that the score of group B decreased more significantly at 6 months and 12 months.

[0138] Statistical results: A set of statistical tables shows the p-values ​​for between-group effects, time effects, and interactions. The interaction p-value < 0.05 indicates that the effects of the two treatment regimens changed differently over time.

[0139] Report: Generate a complete analysis report (PDF) with one click, including methods, results, figures, and a summary of conclusions. The entire analysis process, from queue construction to report generation, takes less than 10 minutes.

[0140] Example 4

[0141] This embodiment provides a research-specific disease database system based on Wilson's disease, including:

[0142] Intelligent follow-up management

[0143] The patient, "Li," was diagnosed by the system as having "hepatic type, Child-Pugh B." Based on its rule engine, the system automatically generated a follow-up plan for him: Month 1: Weekly liver function and coagulation function tests; Months 2-3: Every 2 weeks; After the condition stabilizes: Monthly liver function, complete blood count, and urine copper tests; Liver ultrasound every 3 months.

[0144] The system automatically sends a text message to Mr. Li and his attending physician three days before each follow-up visit.

[0145] At the second month follow-up, the doctor entered the following data: ALT 150 U / L (significantly higher than the previous 80 U / L). The system triggered an "abnormal liver function" warning, with a red indicator reminding the doctor to check immediately. The doctor promptly contacted the patient and discovered that the patient had reduced their medication dosage on their own. After being urged to adjust the treatment plan, serious consequences were avoided.

[0146] All follow-up data were automatically timestamped and linked to Li's case, forming a clear timeline of disease progression.

[0147] After deploying and implementing the system of this invention at our clinical research center, and after six months of operation, the following quantitative results have been achieved:

[0148] Data quality: Data completeness improved from 78% before going live to 99.5%; data accuracy improved from 85% to 99.8%; and data standardization reached 100%.

[0149] Research efficiency: The average time for researchers to screen target patient cohorts has been reduced from >4 hours to <30 seconds; the average time to complete a moderately complex statistical analysis (such as survival analysis or efficacy comparison) has been reduced from 1-2 weeks to 2-4 hours.

[0150] Follow-up management: Patient follow-up compliance rate increased from 70% to 95%; the rate of missing follow-up data decreased from 25% to 2%.

[0151] Research output: Based on the high-quality data and analysis capabilities provided by the system, the center has launched three new clinical studies within six months and efficiently completed the post-analysis of data from two previous studies, significantly accelerating the progress of paper writing.

[0152] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A research-specific disease database system based on Wilson's disease, characterized in that, It includes a case data collection module, a database management module, a clinical research analysis module, a follow-up management module, and a standardized quality control module. The case data acquisition module obtains structured core indicators of copper metabolism, clinical manifestations, genetic information and treatment data based on disease-specific templates; The database management module constructs a knowledge graph containing semantic relationships between gene mutations, clinical phenotypes, treatment plans, and detection indicators, enabling unified storage and efficient retrieval of multi-source data. The clinical research analysis module analyzes copper metabolism dynamics, non-invasive assessment of liver fibrosis, neurological efficacy, and genotype-phenotype relationships by embedding specialized analysis models. The follow-up management module automatically generates and dynamically adjusts personalized follow-up plans based on patient classification, severity, and treatment plan. The standardized quality control module performs logical, reasonable, and complete checks based on knowledge graph rules, and continuously manages the quality of the data.

2. The system according to claim 1, characterized in that, The case data acquisition module interfaces with the hospital information system, laboratory information system and image archiving system to obtain data such as serum ceruloplasmin, 24-hour urinary copper, liver copper content and UWDRS score according to preset field mapping rules, and performs speech transcription and standardization processing on unstructured text information.

3. The system according to claim 1, characterized in that, The database management module includes: The entity recognition unit is used to extract entities such as gene mutations, clinical phenotypes, and detection indicators from clinical texts, literature, and medical records based on the BERT-BiLSTM-CRF model. A relation extraction unit is used to determine the semantic relationships between the entities using a Span-Pair classification model. The fusion processing unit is used to align synonymous entities from different sources based on the edit distance algorithm and the Wu-Palmer semantic similarity fusion rule, and generate a unified node identifier.

4. The system according to claim 3, characterized in that, The database management module uses the Neo4j graph database to store the knowledge graph structure, defining each entity as a node and semantic relationships as edges. The relationship weight is determined by calculating the co-occurrence frequency between nodes, and the priority path is sorted in multi-hop path query according to the weight. The system evaluates the credibility of the query results based on weight thresholds and indicates the strength of the relationship using a color gradient in the results visualization interface.

5. The system according to claim 1, characterized in that, The clinical research analysis module includes a copper metabolism kinetics analysis submodule. This submodule fits time-series data of serum ceruloplasmin and urinary copper concentrations to a nonlinear mixed-effects model. Based on the individualized parameter estimation model, the synthesis rate constant and the clearance rate constant are used to calculate the time point at which steady state is reached using the maximum likelihood algorithm. The model results were then correlated with liver function indicators and neurological symptom scores.

6. The system according to claim 6, characterized in that, The clinical research analysis module further includes a sub-module for evaluating the efficacy of treatment in the nervous system. This submodule analyzes the impact of different copper-chelating drug treatment regimens on the trend of UWDRS score changes using a linear mixed-effects model. Interactive items are established based on patient treatment group, time point, and baseline score, and a visual efficacy trend chart and statistical report are automatically generated after the analysis is completed.

7. The system according to claim 1, characterized in that, The follow-up management module includes a rule base, an inference engine, and an execution unit. The rule base defines follow-up conditions and check items in Boolean logic form, and the inference engine matches patient classification, UWDRS score and treatment plan based on the Rete algorithm. When the monitoring data meets the preset trigger conditions, the execution unit calls the update function to automatically adjust the follow-up interval and detection items, and writes the update record into the database.

8. The system according to claim 1, characterized in that, The standardized quality control module includes a rule definition unit, a verification execution unit, and a quality reporting unit. The rule definition unit sets three types of rules: logical consistency, reasonableness, and completeness, which are used to judge the logical consistency between fields, the reasonableness of the numerical range, and the completeness of key data, respectively. The verification execution unit performs rule verification during the data entry stage and the periodic batch detection stage, and generates an anomaly report with location information. The quality reporting unit statistically analyzes the completeness, consistency, and reasonableness of data, and updates the rule base based on the PDCA cycle mechanism to achieve continuous improvement.

Images